Place and Neighborhood Crime: Examining the
Relationship between Schools, Churches, and
Alcohol Related Establishments and Crime
Prepared by:
Dale Willits
Lisa Broidy
Ashley Gonzales
Kristine Denman
New Mexico Statistical Analysis Center
Dr. Lisa Broidy, Director
March 2011
TABLE OF CONTENTS
TABLE OF CONTENTS.................................................................................................... 1 LIST OF TABLES .............................................................................................................. 2 CHAPTER I: INTRODUCTION ........................................................................................ 3 CHAPTER II: LITERATURE REVIEW ........................................................................... 5 Routine Activity Theory ................................................................................................. 5 Social Disorganization Theory ....................................................................................... 6 Schools and Neighborhood Crime .................................................................................. 7 Churches and Neighborhood Crime................................................................................ 9 Alcohol Establishments and Neighborhood Crime......................................................... 9 Data ............................................................................................................................... 12 Methods......................................................................................................................... 17 CHAPTER IV: RESULTS ................................................................................................ 20 Place and Violent Crime ........................................................................................... 21 Place and Property Crime ......................................................................................... 23 Place and Narcotics Crime ........................................................................................ 25 CHAPTER V: DISCUSSION AND CONCLUSION ...................................................... 28 REFERENCES ................................................................................................................. 34 APPENDIX: ADDITIONAL REGRESSION RESULTS ............................................... 38 1
LIST OF TABLES
Table 3.1. Crime Incidents from 2000 to 2005 by Block Group .........................................13 Table 3.2. Principal Components Matrix of Census Variables Using Varimax Rotation ...15 Table 3.3. Descriptive Statistics for Independent Variables (n = 430 block groups) ..........16 Table 3.4. Moran's I Results .................................................................................................18 Table 4.1. Regression Results on Violent Crime ..................................................................22 Table 4.2. Regression Results on Property Crime ................................................................24 Table 4.3. Regression Results on Narcotics Incidents ..........................................................26 Table A1. Poisson Regression Results Using Homicide as the Dependent Variable ..........38 Table A2. Negative Binomial Regression Results Using Rape as the Dependent Variable39 Table A3. Negative Binomial Regression Results Using Aggravated Assault as the
Dependent Variable ..............................................................................................................40 Table A4. Negative Binomial Regression Results Using Robbery as the Dependent
Variable .................................................................................................................................41 Table A5. Negative Binomial Regression Results Using Burglary as the Dependent
Variable .................................................................................................................................42 Table A6. Negative Binomial Regression Results Using Larceny as the Dependent
Variable .................................................................................................................................43 Table A7. Negative Binomial Regression Results Using Motor Vehicle Theft as the
Dependent Variable ..............................................................................................................44 Table A8. Negative Binomial Regression Results Using Narcotics Incidents as the
Dependent Variable ..............................................................................................................45 2
CHAPTER I: INTRODUCTION
The objective of this research is to determine the degree to which neighborhood crime
patterns are influenced by the spatial distribution of three types of places: schools,
alcohol establishments, and churches. A substantial body of research has examined the
relationship between places and crime. Empirically, this research indicates that there is
more crime at certain types of places than at others (Sherman, Gartin, and Buerger, 1989;
Spelmen, 1995; Block and Block, 1995). The criminological literature also provides
several potential theoretical explanations for these patterns. The routine activity
perspective (Cohen and Felson, 1979) argues that crime occurs when motivated offenders
converge with potential victims in unguarded areas. Places that promote this
convergence are expected to have elevated crime rates, while places that prevent or
reduce this convergence are expected to have lower crime rates. The social
disorganization perspective (Shaw and McKay, 1942; Bursik, 1988; Krivo and Peterson,
1996) argues that communities with more collective efficacy (in the form of internal
social networks and access to external resources and values) are likely to have less crime,
while communities lacking in efficacy are likely to have more crime. Places that promote
the formation of positive social ties and grant the community access to external resources
are expected to reduce crime, while places that inhibit positive social ties and separate the
community from external resources are likely to increase crime.
Much of the literature on place and crime has focused on the influence of bars on
neighborhood crime rates, with a substantial body of research indicating that bars are
associated with elevated crime rates (Roncek and Bell, 1981; Roncek and Pravatiner,
1989; Sherman, Gartin, and Buerger, 1989; Roncek and Maier, 1991; Block and Block,
1995). Sherman, Gartin, and Buerger (1989), for example, found that bars can account
for upwards of 50% of police service calls in a given area. Here we examine the
relationship not only between bars and crime rates, but other types of liquor
establishments as well (e.g., liquor stores and restaurants that serve alcohol). In addition
to the literature that characterizes bars as hot spots for crime, a smaller, yet growing,
body of literature indicates that the presence of schools (Roncek and Lobosco, 1983;
Roncek and Faggiani, 1985; Roman, 2004; Kautt and Roncek, 2007, Broidy, Willits, and
Denman, 2009, Murray and Swatt, 2010) is also associated with neighborhood crime.
The most recent of this research suggests that while high schools are associated with
increased crime at the neighborhood level, elementary schools may have a protective
influence. Research on churches and crime is limited relative to research focused on
schools and bars, but suggests that churches may help protect neighborhoods from crime
(Lee, 2006; Lee 2008; Lee 2010). Furthermore, there are theoretical reasons to suspect
that churches, like schools and liquor establishments, may be an important type of place
to consider when examining crime at the neighborhood level. The current research
contributes to a criminological understanding of place and crime by examining whether
and how all three location types operate to influence crime rates both independently and
relative to one another.
3
Additionally, the current research controls for known neighborhood-level correlates of
crime not often included in research examining place and crime. The place and crime
research has generally been conducted at the smallest geographic unit defined by the U.S.
Census Bureau: the block (for an exception, see Broidy, Willits, and Denman, 2009).
The U.S. Census Bureau, however, releases data on a wider range of social indicators at
larger levels of analysis (like the block group and tract). Consequently, previous research
has been unable to control for a wide array of social-structural factors when examining
the relationship between the distribution of places and the effect of this distribution on
neighborhood crime. This leaves open the question of whether the relationship between
place and crime is independent of factors such as structural disadvantage, residential
mobility, and family disruption or simply a spurious reflection of these established
relationships. In this study, using incident-crime data from Albuquerque, New Mexico,
we assess the influence of place on neighborhood crime rates, net of key structural
correlates of crime. We do this by utilizing the block group as our level of analysis. This
allows us to investigate the effects of schools, liquor establishments and churches, while
controlling for a wider array of variables than previous studies. By controlling for
concepts like structural disadvantage, residential mobility, and family disruption, we can
be more certain that any significant relationship between place and neighborhood crime is
reflective of place effects and not of structural conditions.
In sum, the current research examines the relationship between place and crime. We
focus specifically on the relationship between schools, churches, and bars (including
other alcohol establishments) and neighborhood crime. We focus particularly on the
relative influence of different types of places on crime rates and on variations in the
influence of place type by crime. In other words, do some place types have a more
consistent influence on crime than others and does the strength of that influence depend
on the type of crime we examine (e.g., violent v. property crime)?
This report is divided into five chapters. The second chapter describes the theoretical
framework utilized in this research and reviews the relevant literature on place and
neighborhood crime. The third chapter describes the data and methodologies that we
used to investigate the relationship between place and crime. The fourth chapter presents
the results of our research. The final chapter discusses these results, presents empirical
and theoretical conclusions, and addresses directions for policy and future research.
4
CHAPTER II: LITERATURE REVIEW
Research examining the relationship between place and neighborhood crime is typically
framed in terms of routine activities theory and social disorganization theory. In the
following literature review we briefly describe each of these theoretical traditions. Then,
we discuss how these theories might explain the relationship between specific types of
places and neighborhood crime.
Routine Activity Theory
Routine activity theory states that criminal acts require the convergence of three
elements: motivated offenders, suitable targets, and the absence of capable guardianship
(Cohen and Felson, 1979: 589). Building on utilitarian principles, the theory assumes
that the absence of any one of these elements increases the risks of crime relative to its
rewards, making crime less likely. At the macro-level, routine activity theory is rooted in
the social and physical ecology perspective. According to this perspective, crime is
“affected not only by the absolute size of the supply of offenders, targets, or
guardianship, but also by the factors affecting the frequency of their convergence in space
and time” (Sherman, Gartin, and Buerger, 1989: 30-31). In other words, specific places
are likely to be crime prone because of associated routine activity patterns that are
conducive to crime.
A specific place may be criminogenic because it increases the prevalence of any
combination of the three elements described by the routine activity perspective. The
routine activity argument, when applied to types of places, focuses on the opportunity
structures associated with specific places and how these shape the related probability that
a criminal event occurs. Places that are frequented by motivated offenders are likely to
exhibit a higher incidence rate of crime than others, as an increase in the number of
would-be offenders present at a specific place increases the probability that these
potential offenders will converge in time and space with suitable victims. Similarly,
locations frequented by suitable victims are also expected to be more criminogenic than
places that are avoided by suitable victims. Places where both potential offenders and
potential victims converge in large numbers are expected to be especially criminogenic.
Increases in the presence of would-be offenders and/or would-be victims are also
expected to decrease the effectiveness of guardianship, as larger crowds skew the
guardian-to-person ration, making it harder for would-be guardians to comprehensively
monitor all of the people present in a given place. From this perspective, locations that
promote the convergence of at-risk individuals or groups (that is, individuals or groups
that can be generalized as “motivated offenders” or “suitable targets” because of their
elevated levels of offending and victimization) are expected to be more criminogenic than
others.
5
While Cohen and Felson (1979) did not address the issue of criminal motivation, other
researchers employing the routine activities perspective have also argued that places can
increase both the motivation of potential offenders and the suitability of potential victims.
Roncek, in a series of papers with various co-authors (Roncek and Bell, 1981; Roncek
and Pravatiner, 1989; Roncek and Maier, 1991) argued that the consumption of alcohol
that occurs at bars increases both offender motivation and target suitability. Offender
motivation rises because alcohol decreases inhibition and increases aggression. At the
same time, target suitability increases because inebriation may decrease potential victims’
ability to guard themselves and their belongings.
Social Disorganization Theory
Social disorganization theory (Shaw and McKay, 1942; Sampson and Groves, 1989;
Sampson and Wilson, 1995; Sampson, Raudenbush, and Earls, 1997) states that
disorganized neighborhoods lack the capacity to self-regulate and organize against
criminal behavior and, as such, have higher crime rates than other neighborhoods. These
socially disorganized neighborhoods are generally characterized by structural
disadvantage (Bursik, 1988) and sparse conventional social networks (Kasarda and
Janowitz, 1974), which impede collective efficacy. Without collective efficacy,
community members do not work together to maintain the types of cohesive community
ties that deflect crime. A wide body of research called the “neighborhood effects”
literature has developed around and produced considerable empirical support for social
disorganization theory (Sampson, Morenoff, and Gannon-Rowley, 2002)
Using this social disorganization framework, neighborhood scholars have attempted to
specify the processes through which community organization influences collective
efficacy and informal social control. Building from Sampson and Wilson’s (1995)
arguments regarding social isolation, some researchers have focused on the role of local
institutions within socially disorganized communities. Krivo and Peterson (1996), note
that:
“disadvantaged communities do not have the internal resources to organize
peacekeeping activities… and at the same time, local organizations (churches,
schools, recreation centers) that link individuals to wider institutions and foster
mainstream values are lacking” (1996: 622).
Krivo and Peterson suggest that strong local institutions can mitigate the effects of
structural disadvantage. One implication of this statement is that they expect local
places, like schools, to be associated with social disorganization above and beyond
structural predictors. This is both because these places provide tangible resources (in the
form of organizational opportunities) to the neighborhood and because these places
reduce a neighborhood’s level of social isolation. For example, neighborhoods with
schools are more likely to contain parent-teacher associations and the presence of these
organizations may foster social ties in the neighborhoods and reduce the social isolation
of the neighborhood (via the fostering of involvement from parents and teachers that have
6
resources that extend beyond the immediate area of the school’s grounds). There is some
evidence to support this hypothesis, as Peterson, Krivo, and Harris (2000) reported a
modest reduction in neighborhood crime rates when community centers are present.
Other research indicates that elementary schools may reduce crime at the neighborhood
level (Broidy, Willits, and Denman, 2009; Murray and Swatt, 2010).
In general, social disorganization theory states that places that promote social
organization and foster positive social ties should reduce neighborhood crime, while
places that hinder social organization should increase neighborhood crime.
Schools and Neighborhood Crime
The relationship between schools and crime seems to vary by type of school. A small
body of research has demonstrated that there is a relationship between schools and crime
(Roncek and Lobosco, 1983; Roncek and Faggiani, 1985; Roman, 2004; Kautt and
Roncek, 2007, Broidy, Willits, and Denman, 2009, Murray and Swatt, 2010), though all
of the early research on this topic narrowly focused on high schools. This is sensible as
there are several reasons to expect high schools to be associated with higher levels of
neighborhood crime. High schools are populated by an age-group that is known to both
commit higher rates of crime and experience higher rates of victimization. High schoolaged youths are more likely to be offenders than individuals in any other age group with
the exception of young adults. The relationship between age and crime is widely
accepted among criminologists and appears to hold true across race, gender, society, and
time (Hirschi and Gottfredson, 1983; Farrington, 1986). Therefore, motivated offenders
gather in and around schools on a daily basis. Moreover, given that youth are at an
increased risk for criminal victimization (Rand and Catalano, 2007), it is clear that
suitable targets also gather in and around schools. High schools and to a lesser degree,
middle schools, therefore, bring together large groups of individuals from age groups that
are characterized by higher offending and victimization rates.
Furthermore, research indicates that a substantial proportion of youth victimization is
related to the routine activities of attending school (Garofalo, Siegel, and Laub, 1987).
Teacher student ratios in many schools are such that capable guardianship is often absent,
a situation that is compounded in larger schools. This means that youth convene in and
around schools with limited adult guardianship. Given the convergence of motivated
offenders and suitable targets in the school environment, these limitations on capable
guardianship should further increase crime and victimization at or near schools.
These arguments are not readily applicable to elementary schools, as elementary school
students are less likely to be both offenders and victims. Moreover, elementary schools
have smaller populations and smaller student to teacher ratios, which might indicate that
there is more capable guardianship in and around elementary schools than at middle and
high schools. From a social disorganization perspective, elementary schools might
actually be thought to increase the social organization of a neighborhood.
7
Schools and their related activities, organizations, and events facilitate social ties among
both adults and adolescents. Moreover, schools promote the formation of local
organizations, like parent-teacher associations, and add additional structure and
supervision to the juvenile population, both through the process of schooling and through
associated extracurricular clubs and activities. Indeed, social disorganization theorists
have argued that the local organizations and youth supervision are important aspects of
maintaining community organization (Shaw and McKay, 1942; Sampson and Groves,
1989). It may be the case that elementary schools are particularly effective at promoting
this kind of social organization since they are generally smaller, fostering a more closeknit school community. In addition, parents tend to be more involved in their children’s
education and related school activities during the elementary school years (Hill and
Taylor, 2004; Eccles and Harold. 1996). While few studies have addressed the role of
elementary schools in neighborhood crime, there is some evidence that elementary
schools may, in fact, be a protective factor at the neighborhood level (Broidy, Willits, and
Denman, 2009, Murray and Swatt, 2010; though see Kautt and Roncek, 2007 for
contrary evidence).
Conversely, by high school, parents are notably less involved in their children’s
education. At this stage, adolescents are becoming more autonomous and the school
curriculum becomes more advanced, so students are less likely to seek parental
involvement and parents feel less qualified to offer academic help (Hill and Taylor, 2004;
Eccles and Harold, 1996). Shaw and McKay (1942) noted that among the characteristics
of socially disorganized areas is the presence of groups of unsupervised adolescents. In
that sense, high schools and middle schools may actually contribute to a neighborhood’s
social disorganization. The routine activity and social disorganization perspectives
overlap considerably on this issue, as both traditions argue that groups of un- (or under-)
supervised youths are a criminogenic risk factor for neighborhoods. The research on
schools and neighborhood crime supports these arguments, as a number of studies have
found that neighborhoods with middle schools (Roman, 2004; Broidy, Willits, and
Denman, 2009) and high schools (Roncek and Lobosco, 1983; Roncek and Faggiani,
1985; Roman, 2004; Broidy, Willits, and Denman, 2009, Murray and Swatt, 2010) have
higher crime rates than neighborhoods without middle or high schools.
Given the previous research on schools and crime and the theoretical rationale from the
routine activities and social disorganization perspectives, we make the following
hypotheses regarding the relationship between schools and neighborhood crime.
H1: Neighborhoods containing high schools or middle schools will have more
crime than neighborhoods without high schools or middle schools, controlling for
other factors.
H2: Neighborhoods containing elementary schools will have less crime than
neighborhoods without elementary schools, controlling for other factors.
8
Churches and Neighborhood Crime
Research on the relationship between churches and neighborhood crime is limited.
Instead, research has focused on the individual-level effects that religious beliefs or
religious affiliation have on criminal behavior (Alvarez-Rivera and Fox 2010; Johnson,
Jang, De Li and Larson 2000; Knudten 1971). However, a small body of empirical
literature has demonstrated that there is a negative relationship between churches and
crime in rural counties (Lee, 2006; Lee, 2008; Lee, 2010).
While empirical research on churches and neighborhood crime is rare, theorists have
argued that churches may factor into neighborhood crime. In particular, social
disorganization theorists have cited churches as being strong local institutions that could
create social ties and thereby hinder or prevent crime (Rose and Clear 2006; Krivo and
Peterson 1996). Social disorganization theorists argue that similar to schools, churches
could contribute to and be reflective of a community’s level of collective efficacy.
Similarly, routine activity theorists might argue that churches decrease crime at the
neighborhood level. First, in order for churches to be considered criminogenic hot spots,
motivated offenders must frequent or be near or around the church. Notably, churches
have by definition been places of social support, meditation, healing and religious unity,
and “services offered by churches and their members include providing a therapeutic
haven that buffers the impact of psychological distress and material/financial needs”
(Taylor and Chatters 1988; Wimberly 1979). The assumption is that people seek out
churches for religious guidance and would not be at or near churches actively seeking out
the opportunity to commit crime (Taylor, Thornton and Chatters, 1987). Although
churches do bring congregations of people together, including potentially suitable targets
and motivated offenders, it may be the case that many church attendants within a specific
congregation know each other and would be able to effectively “self-police” and protect
each other from crime. This “self-policing,” or informal social control, would serve to
increase the risks associated with committing crime and decrease the incentives, reducing
the overall likelihood of criminal activity (Cohen and Felson, 1979). Therefore, both the
routine activity and social disorganization perspectives highlight a potentially protective
role for churches in neighborhood crime. Thus far, empirical research has not addressed
these arguments.
H3: Neighborhoods with churches present will have lower levels of crime than
neighborhoods without churches.
Alcohol Establishments and Neighborhood Crime
Previous research utilizing the routine activity theory as a framework has linked certain
location types to crime, designating these locations as hot spots. A number of studies, for
example, have identified bars as criminogenic locations ((Roncek and Bell, 1981; Roncek
9
and Pravatiner, 1989; Sherman, Gartin, and Buerger, 1989; Roncek and Maier, 1991;
Block and Block, 1995). While research has not addressed the role of bars in the
formation of social organization, it is possible that alcohol-related establishments
increase neighborhood crime by decreasing social organization. These establishments are
likely frequented by people from both inside and outside of the neighborhood in which
they are located. The outsiders are likely to be less connected to the neighborhood in
question and therefore are less apt to be deterred from crime by neighborhood-level
informal social control. Even for insiders, the inhibition-reducing effects of alcohol may
diminish the strength that informal social controls generally hold over behavior. That
bars generate a disproportionate number of police service calls relative to other
neighborhood places suggests that the social processes that inhibit crime in other
neighborhood places are less effective at or around bars.
From the routine activities perspective, these locations are criminogenic because they
provide an opportunity for the intersection of offenders and targets in the absence of
guardianship. Bars, for example, are occupied by individuals who “are likely to have
cash with them and thereby present opportunities for crime, especially if they become
intoxicated” (Roncek and Maier, 1991: 726). In other words, bar patrons carry cash,
which makes them a suitable target for property offenses, and may be less capable of
guarding themselves and their assets when intoxicated. A substantial body of research
links alcohol consumption to aggression (Ito, Miller, and Pollock, 1996). In addition,
bars can be activity hubs, where larger groups of individuals congregate. The density of
individuals at bars can increase the likelihood of crime by increasing anonymity and
reducing the effects of supervision and guardianship (Roncek, 1981). Block and Block
(1995) build on the density argument, noting that bars and dance clubs that are clustered
in night-life areas are more likely to be hotspots than isolated bars and dance clubs.
It is important to note that the routine activity perspective argues that hot spots will
generate crime both at the specific hot spot place and in the surrounding area. The hot
spot, itself, is expected to generate crime due to the convergence of motivated offenders
and suitable targets. The areas directly surrounding the hot spots are also expected to
generate crime for the same reasons, as the areas around the hotspots will contain the
routes to and from the hot spot. Moreover, the hot spots themselves may be more
supervised than areas directly surrounding the hot spot. Bars, for example, may have
security guards and bouncers that prevent crime within the actual building. The areas
directly surrounding bars, however, may not be as closely supervised.
Other alcohol establishments may be related to crime for similar reasons. For example,
both restaurants that serve alcohol and social clubs (e.g., the Eagles or Elks) where
alcohol is served may generate additional crime in a neighborhood for similar reasons.
Both restaurants and social clubs promote the convergence of people in time and space,
which from a purely probabilistic standpoint should increase the likelihood of criminal
events. This likelihood may be further bolstered by the fact that alcohol is served at these
places and, as previously mentioned, alcohol may work toward both increasing the
motivation of and the vulnerability to offending. However, the strength of the
10
relationship between these types of places and crime may be weaker than the relationship
between bars and crime. In terms of restaurants and bars, it may be the case that a greater
proportion of people visit bars with the intent of consuming alcohol and engaging in
social interactions with strangers, while people at a restaurant may be more focused on
eating and socializing with the people at their table. Similarly, the people who attend
social clubs share some common group membership and may be less inclined to view
each other as suitable targets. In this way, it is possible that these types of places may
increase the routine activity processes that promote crime but also the collective efficacy
that inhibits it.
H4: Neighborhoods containing alcohol-selling or -serving establishments will
have more crime than neighborhoods without these types of places, controlling for
other factors.
H5: The criminogenic effect of bars will be greater than the criminogenic effect
of other liquor establishments.
In addition to examining place specific hypotheses, we also explore the relative
contributions of each place type to neighborhood crime. By doing so, we evaluate which
types of places are more or less strongly related to specific types of crime. The literature
offers little theoretical or empirical guidance regarding the relative influence of specific
place types. As such, we do not propose any specific comparative hypotheses but rather
offer an exploratory comparative analysis that begins to assess the relative influence of
distinct place types on varying crime types. We believe that the results of this
comparative analysis may further inform theory, research, and policy on place and crime.
Indeed, neighborhoods invariably include a diverse array of place types whose relative
contributions to crime are important to consider when weighing the costs and benefits of
potential policy initiatives.
11
CHAPTER III: RESEARCH DESIGN
In order to address the research questions and hypotheses described in the previous
section, we estimated several regression models using block group level data. This
section of the report describes the data and methods utilized in this research.
One of the key complications associated with studying neighborhood crime is that it is
difficult to empirically specify the boundary of a neighborhood. Most neighborhood
research on crime has used U.S. Census-defined jurisdictions, like census tracts, block
groups, and blocks, to approximate neighborhoods and neighborhood patterns and trends
(Sampson, Morenoff, and Gannon-Rowley, 2002). While Census designations are
artificial and are not necessarily reflective of the lived experience of being in a
neighborhood and/or community, they are often the only option for researchers interested
in meso-level processes, as they are easily identifiable and are connected to a wide range
of data collected by the U.S. Census Bureau.
Indeed, both the schools and neighborhood crime and the bars and neighborhood crime
literatures have almost exclusively used census designations as the unit of analysis. The
block is the smallest of the census designations and has been frequently used to determine
if schools or bars produce crime in the areas directly surrounding their location. In order
to account for the possibility that places generate crime in surrounding areas, researchers
studying place and neighborhood crime often construct adjacency measures to account
for blocks that are near blocks with schools or bars.
For the current research, however, we have opted to utilize the block group level of
analysis. Block groups are the second smallest census designation and are made up of
blocks. While utilizing blocks would make the current research more directly
comparable to previous research on place and neighborhood crime, doing so would limit
our ability to control for the potential influence of social factors. The U.S. Census
Bureau maintains more social, economic, and demographic information at the block
group level than it does at the block level. In particular, a variety of measures of
structural disadvantage (including measures of education, income, and employment) are
available only at the block group and larger levels of aggregation. By utilizing the block
group level of analysis, we are able to determine if the relationship between place and
crime is reflective of some actual trend and not spurious by way of social structural
factors.
Data
The data for this research cover three areas: crime, social and demographic features of
neighborhoods, and places.
The incident-level crime data used in this report were provided by the Albuquerque
Police Department (APD). These data cover the years 2000-2005 and include the date,
12
time, location, and crime code and statute violation for each documented incident. For
this project, we included the following offenses: homicide, rape, robbery, aggravated
assault, burglary, larceny, motor vehicle theft, and narcotics violations. By including the
part 1 violent and property offenses as well as narcotics violations, we are able to
investigate the link between place and a variety of different forms of crime at the
neighborhood level. While it is reasonable to assume that certain types of places, like
schools, might be more strongly related to other less serious crime types (like, for
example, vandalism and simple assault), we opted to focus on serious crimes, as it is
generally assumed that serious offenses are more often reported to and accurately counted
by the police (Mosher, Miethe, and Phillips, 2002).
In order to examine neighborhood crime patterns, crime incidents from 2000 to 2005
were geocoded using ArcGIS mapping software. Once mapped, incidents were matched
to census block groups and aggregated, providing a sum of crime incidents over that time
period within each census block group in Albuquerque. These counts were summed from
2000 to 2005 to account for yearly fluctuations and to induce additional variation
(thereby improving our ability to account for variance in crime across block groups).
These data were then exported to SPSS and linked to block group social and demographic
data and to school data. A summary of the crime data included in this analysis is
presented in Table 3.1.
Table 3.1. Crime Incidents from 2000 to 2005 by Block Group
Mean
Part 1 Violent
Part 1 Property
Homicide
Rape
Robbery
Aggravated
Assault
Burglary
Larceny
Motor Vehicle
Theft
Narcotics
Incidents
38.32
234.41
0.37
3.47
9.05
Standard
Deviation
41.47
308.82
0.71
4.37
11.86
Minimum
Maximum
0
0
0
0
0
396
2823
4
41
80
25.43
27.84
0
279
45.18
155.85
39.07
258.21
0
0
352
2615
33.37
35.56
0
298
27.03
57.40
0
952
The social and demographic data for block groups were compiled from the Census 2000
Summary File 3. As previously mentioned, the U.S. Census Bureau provides
substantially more information about block groups than blocks. In order to account for
social disorganization explanations, we obtained data on a number of structural
disadvantage and mobility variables, including: the percentage of renter-occupied
housing in a block group, the percentage of households with children headed by single
13
females in a block group, the percentage of a block group that is unmarried, the
percentage of the block group that has moved in the last 5 years, the percentage of
housing that is vacant in a block group, the percentage of people with less than a high
school education in a block group, the percentage of people living under the poverty line
in a block group, the percentage of households in a block group receiving public
assistance, and the joblessness (employed individuals plus those not in the labor market)
in a block group.
As suggested by previous research on social disorganization, many of these variables are
collinear (Sampson, Raudenbush, and Earls, 1997). Accordingly, we were unable to use
all of these variables separately in our analysis. In order to address this collinearity, we
utilized Principal Components Analysis (PCA) on this list of variables. PCA is a data
reduction technique, which when performed on a matrix of variables, produces
uncorrelated components that account for the shared correlation and/or covariance of the
variables included in the analysis (for details, see Dunteman, 1989). These components
can be calculated as standardized scores, indicating whether a specific observation scores
low, average, or high on that particular component.
For the current research, we utilized SPSS to conduct PCA on the list of variables above.
This procedure, using a varimax rotation to improve interpretation, produced two
components with eigenvalues greater than 1, which together, accounted for nearly 70% of
the variance in the variables. The eigenvalues of the first two components, along with
associated scree charts, allowed us to exclude the remaining components on the grounds
that the first two components adequately address the variance in included variables.
The results of the PCA are listed below in table 3.2. This table, which is a principal
components matrix of census variables, lists the correlation between each variable on the
components produced from the PCA. The variables percentage of the population with
less than a high school diploma, percentage of the population jobless, percentage of
households living under the poverty line, median income, and percentage of households
receiving public assistance loaded in the first principal component. We call this
component score structural disadvantage. The variables percentage of people that have
moved in the last 5 years, percentage of housing vacant, percentage of dwellings
occupied by renters, and percentage of households with children headed by single
females loaded on the second principal component. We named this component score
instability.
In addition to the variables described above, we also gathered information on the total
population of block groups, the percentage of the population that is Hispanic, and the
percentage of the population 18 and under from the 2000 Census. Preliminary analysis of
these variables suggested that they were not collinear with the instability and
disadvantage measures described above, and thus they were maintained as separate
independent variables.
14
In addition to the data representing social-structural conditions, we also constructed
variables to indicate whether or not a specific type of place was present within each block
group. The school-presence data came from the National Center for Education Statistics
(NCES). Using the NCES website, we obtained a list of public schools that were open in
Albuquerque from 2000 to 2005. Using the street addresses for these schools (which
were also available via NCES), we geocoded each school and associated it with a specific
block group.
Table 3.2. Principal Components Matrix of Census Variables Using Varimax
Rotation
Variable
% Less than High School
Diploma
% Joblessness
% Poverty
Median Income
% Public Assistance
% Moved in Last 5 Years
% Vacant
% Renters
% Female Headed
Households
Eigenvalue
Disadvantage
0.8995
Instability
0.0169
0.7888
0.8048
-0.6752
0.7823
-0.1783
0.2536
0.2362
0.3741
-0.1105
0.4053
-0.5083
0.1600
0.8427
0.6843
0.8886
0.6062
4.34 (48.24%)
1.89 (21.01%)
Bold indicates which component the variable loaded more strongly on.
As all establishments that serve or sell alcohol are required to obtain a liquor license in
New Mexico, we were able to construct a list of alcohol-severing establishments that
were open from 2000 to 2005 using publically available data from the New Mexico
Regulation & Licensing Department. We categorized the alcohol-serving establishments
as bars, restaurants that serve beer and wine, retail establishments that sell alcohol, and
liquor clubs. We primarily assigned each establishment by the type of state license under
which it operates. The State of New Mexico issues a “restaurant” specific liquor license
for beer and wine. Some restaurants, however, have full service bars. As there was no
way to distinguish restaurants with full bars from bars, these establishments are all coded
as bars. Retail establishments include both convenience stores, liquor stores, grocery
stores, and big box stores that sell alcoholic beverages, while liquor clubs are nonprofit
organizations with liquor licenses (like the Veterans of Foreign Wars).
The addresses for these establishments came from the New Mexico Regulation and &
Licensing Department, when available, and from Google Maps when the NM Regulation
& Licensing Department did not provide a physical address. Using these street
addresses, we geocoded each alcohol-serving establishment and associated it with a
specific block group.
15
And finally, the church addresses were pulled from land use maps maintained by the City
of Albuquerque and were associated with specific block groups using ArcGIS.
We merged all of the census, crime, and place data together in a single dataset. Using
this merged data, we created dummy variables to indicate whether a specific type of place
was present in each block group. These dummy variables are: elementary school (1 if
there is an elementary school within that block group, 0 otherwise), middle school (1 if
there is a middle school within that block group, 0 otherwise), high school (1 if there is a
high school within that block group, 0 otherwise), bar (1 if there is a bar within that block
group, 0 otherwise), restaurant (1 if there is a restaurant within that block group, 0
otherwise), retail establishment (1 if there is a retail establishment within that block
group, 0 otherwise), liquor club (1 if there is an liquor club within that block group, 0
otherwise), and church (1 if there is a church within that block group, 0 otherwise). We
tested other formulations of these variables (including raw counts), but found that the
dummy variable specifications were more appropriate for our analysis, as certain block
groups were outliers in terms of the volume of bars and churches.
Table 3.3 provides the descriptive statistics for the independent variables included in this
report.
Table 3.3. Descriptive Statistics for Independent Variables (n = 430 block
groups)
Variable
Disadvantage
Instability
% Hispanic
% Under Age
18
Bar
Retail
Liquor Club
Restaurant
Church
Elementary
School
Middle School
High School
Population
Mean
0
0
40.96
24.98
Std. Deviation
1
1
24.05
8.33
Minimum
-1.77
-2.05
0
3.34
Maximum
3.62
3.57
100
71.99
0.22
0.24
0.06
0.27
0.47
0.17
0.42
0.43
0.24
0.56
0.50
0.38
0
0
0
0
0
0
1
1
1
1
1
1
0.05
0.03
1294.60
0.23
0.17
646.38
0
0
31
1
1
4355
The variables disadvantage and instability have means of 0 and standard deviations of 1
because they are standard normal variables that were constructed using a regression
scoring technique based on the previously described principal component analysis. The
means for the place variables indicate the percentage of block groups that contain a
specific place (for example, 22% of block groups in Albuquerque contain a bar).
16
Methods
We utilize regression techniques to determine the relationship between place and crime at
the block group level. Note, however, that because criminal incidents are discrete events
and because many of the crime types covered in this analysis are heavily skewed to the
right, traditional ordinary least squares regression techniques are inappropriate. Poisson
regression, a variant of a generalized linear model, is typically preferred to ordinary least
squares when dealing with count data (Osgood, 2000). Poisson regression models
describe the relationship between a set of independent variables and the expected count
of a dependent variable.
The Poisson regression model, however, assumes that the mean is equal to the standard
deviation. This is not the case for these data since all of the crime variables included in
this analysis, with the exception of homicide, are over-dispersed (that is, they have a
variance significantly greater than the mean). In these cases, it is common to utilize
negative binomial regression (Osgood, 2000). Negative binomial regression possesses
qualities similar to Poisson regression, while including an extra regression coefficient to
account for overdispersion. Specifically, negative binomial regression maintains the
same e to the b style of interpretation as Poisson regression. In the results chapter, the
regression results dealing with homicide utilize Poisson regression, while the regression
results dealing with all other dependent variables utilize negative binomial regression.
All regression models were estimated using STATA software.
Poisson and negative binomial regressions are utilized when the dependent variable is
discrete. In this case, the dependent variables are crime counts. However, it is likely that
block groups with more people have more crime. In order to account for this possibility,
we control for population in our regression models. Instead of including population as a
normal independent variable, which would suggest that population has a direct and
substantively interesting relationship with crime counts, we include population as an
exposure variable. The natural logarithm exposure variable is entered into the right hand
side of Poisson or negative binomial regression and is given a fixed coefficient of one,
which essentially changes Poisson and negative binomial regression from an analysis of
crime counts to an analysis of crime rates per capita (Osgood, 2002: 27).
We also addressed spatial dependency in each of the regression models presented in the
results section of this paper. Spatial dependency is the idea that geographically close
units are likely to be more similar to each other than to units that are geographically
distant. Spatial dependency can come from multiple sources, including the artificial
nature of census jurisdiction and “spillover.”1 Significant spatial dependency can lead to
1
Spatial dependency can result from census jurisdictions in that they may not accurately capture
the active units of analysis. For example, suppose crime in a pair of block groups stems from a set of
neighborhood processes and structures. If the block groups cut that neighborhood in half, then each of the
block groups is expected to have a similar count of criminal incidents. Spatial dependency resulting from
spillover suggest that geographic areas affect and are affected by neighboring areas. While conceptually
distinct from the problem of artificial jurisdictions, spillover will also result in block groups that are
17
issues of spatial autocorrelation in statistical procedures. Spatial autocorrelation is a
substantial problem which indicates that the assumptions of independence for a given
regression model are not met. For regression analyses, spatial autocorrelation can result
in unstable regression coefficients and inaccurate standard error estimates. In other
words, it is difficult to determine the effects of independent variables in the presence of
spatial autocorrelation.
Using the CrimeStat Spatial Statistics program, we calculated Moran's I for each of the
dependent variables utilized in our analysis. Moran's I is a common measure of spatial
autocorrelation, which uses a weighted correlation technique to determine if data are
independently distributed across space (Anselin, 1992). As indicated in table 3.4, each of
our dependent variables demonstrated significant clustering (that is, they had I values that
were significantly greater than expected under the assumption of spatial independence).
In order to address this spatial autocorrelation, we used GeoDa software to calculate
spatial lags for each dependent variable in our analysis. The spatial lag is defined as
∑j ωij x j , where x j is the j-th observation of variable x and ωij is the weight from the i-th
row of the spatial weights matrix (Anselin, 1992). This is essentially the weighted
average of values in adjacent block groups. Therefore, spatial lags account for spatial
autocorrelation by controlling for levels of a variable in surrounding areas.
Table 3.4. Moran's I Results
Variable
Moran’s I Value
(Expected:-.0023)
Property Sum
.1498
Homicide
.0926
Violent Sum
.2886
Rape
.2044
Robbery
.1967
Aggravated Assault .3217
Burglary
.3226
Motor Vehicle Theft .3035
Larceny
.1030
Narcotics
.1597
99 Permutations
Significance
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
999 Permutations
Significance
.001
.003
.001
.001
.001
.001
.001
.001
.002
.001
And finally, it should be noted that we do not include an overall adjacency measure for
place in our regression analyses. Many of the previous studies on place and
neighborhood crime have included adjacency measures (Roncek and Lobosco, 1983;
Roncek and Faggiani, 1985; Kautt and Roncek, 2007). Typically, these adjacency
expected to have similar counts of criminal incidents. In both cases, these similarities suggest that crime
may not be independently distributed across geographic units.
18
measures are dummy variables that indicate whether or not an adjacent geographic unit
(always the census block in previous research) contains a given place (e.g., is a block
close to block that contains a school). This measure is typically intended to capture the
effects of places on crime in nearby areas.
We do not include any adjacency measures for two reasons. First, we are utilizing a
larger unit of analysis. As block groups are made up of blocks, significant results in our
analysis suggest that places influence crime in the block group, not just in the specific
block in which they are located. Secondly, at this level of analysis, the vast majority of
blocks are near block groups that contain a specific type of place. For example, 400 of
the 432 block groups in Albuquerque are adjacent to one or more block groups that
contain a school. Similarly, about 50% of block groups in Albuquerque contain a church
and nearly 50% contain an alcohol-related establishment of some sort. In other words, at
larger levels of aggregation, there is not enough variation to warrant the inclusion of an
adjacency measure. Ultimately, this is a trade off for using the block group level of
analysis. The block group allows us to control for more social and economic indicators
and thus to give the social disorganization perspective a more thorough test. However,
because block groups are larger and because the types of places included in our analysis
are spread across Albuquerque, we are unable to test for adjacency effects and therefore
can only make general conclusions about block groups and not about surrounding areas.
19
CHAPTER IV: RESULTS
In order to investigate the relationship between place and crime at the block group level,
we estimated a series of regression models using three dependent variables: violent
crime, property crime, and narcotics incidents.
The dependent variable Violent Crime is the number of the part 1 violent offenses that
occurred in a given block group (that is, the sum of homicide, rape, aggravated assault,
and robbery over 2000 to 2005). The dependent variable Property Crime is the number
of the part 1 property offenses that occurred in a block group (that is, the sum of burglary,
larceny, and motor vehicle theft). The dependent variable Narcotics is the number of
drug possession and distribution arrests that occurred in a block group. The regression
results for each dependent variable are presented in separate tables (Tables 4.1, 4.2, and
4.3) with five models. The five models include a control model (a regression model
containing only control variables), a school model (a regression model containing control
variables and dummy variables for the presence of elementary schools, middle schools,
and high schools), a church model (a regression model containing control variables and
the dummy variable for church presence), an alcohol model (a regression model
including control variables and dummy variables for bars, retail establishments, liquor
clubs, and restaurants), and a complete model (a regression model containing all of the
above variables).2
All of the results presented below utilize negative binomial regression, as all of the
models demonstrated significant overdispersion. We have included both pseudo Rsquared values and negative log-likelihood (-2LL) values to indicate model fit. While the
pseudo R-squared values are more readily interpretable, there are some technical
problems with these statistics. 3 All comparisons of models are done using the loglikelihood ratio test. The test statistic for the log-likelihood ratio test is calculated as
twice the negative difference of the log-likelihood values for two nested models and is
assumed to have a chi-squared distribution with the degrees of freedom equal to the
number of additional variables in the larger of the two nested models. If this test statistic
is greater than a crucial chi-squared value with equal degrees of freedom, then the model
with additional variables is said to improve the fit of the original model.
2
We also estimated a series of five regression models for each distinct crime type, in order to
determine if there were important differences in the relationship between place and specific crime types.
These supplementary regression models also use the negative binomial framework, except for the models
utilizing homicide as the dependent variable. Homicide did not display substantial overdispersion, and thus
we used Poisson regression to analyze homicide. In general these results are substantively similar to the
results presented below. Tables displaying these regression results are included in Appendix A.
3
In some instances, pseudo R-squared measures have a maximum possible value that is less than 1,
suggesting that these measures will be artificially lower than the traditional R-squared measure used in
ordinary least squares regression (Dobson, 2002).
20
The regression coefficients in these tables are unstandardized negative binomial
regression coefficients. As these coefficients are the expected change in the log of the
expected count of y given a unit change in x, these coefficients are somewhat difficult to
interpret in their unstandardized form. In order to facilitate interpretation, these
coefficients are exponentiated prior to interpretation. This process changes the
coefficients to the expected multiplicative change in the expected count of y. In other
words, if the regression coefficient for variable X was 0.350 and this was exponentiated
(e 0.350 = 1.419) , that would indicate that a 1-unit increase in X would correspond to a
41.9% increase in the expected count of Y.
Place and Violent Crime
Table 4.1 displays the regression results for violent crime. The coefficient for the spatial
lag is statistically significant in all five models. This indicates that there is a significant
amount of clustering of violent incidents. The control variables are all statistically
significant as well. Block groups with higher levels of disadvantage and instability are
expected to have significantly more violent crime than more advantaged and stable block
groups. Block groups with larger Hispanic populations are also expected to report
significantly more violent incidents, while block groups with larger youth populations are
expected to report significantly fewer violent incidents.
In terms of specific places, alcohol-related establishments have the largest statistically
significant relationship with violent crime at the block group level. Block groups that
contain bars, alcohol severing establishments, and restaurants that serve beer and wine
are expected to report significantly more violent crime incidents than block groups
without those types of places. The complete model indicates that block groups with bars
are expected to report 53.0% ( e 0.425 = 1.529 ) more violent crime incidents than other
block groups, controlling for other factors. Similarly, block groups with alcohol-selling
establishments are expected to report 46.9% ( e 0.385 = 1.469 ) more violent crime
incidents than block groups without alcohol-selling establishments, and block groups
with restaurants that serve beer and wine are expected to report 30.0% ( e 0.262 = 1.299 )
more violent crime incidents than block groups without these types of restaurants. These
results support hypotheses 4 and 5. Block groups with alcohol-serving and -selling
establishments appear to have significantly more violent crime than block groups without
these establishments. Furthermore, bars increase violent crime at the block group level
more substantially than restaurants or liquor clubs.
Schools and churches appear to be generally unrelated to violent crime at the block group
level. The schools model suggests that block groups with high schools are expected to
report significantly more violent crime incidents than block groups without high schools,
but this relationship is not statistically significant in the complete model. This suggests
that to the degree that there is a relationship between schools and violent crime at the
block group level, this relationship is weaker than the relationship between alcohol21
serving and -selling establishments and violent crime at the block group level. These
results are mildly supportive of hypothesis 1 in that there is some evidence that block
groups with high schools are expected to report significantly more crime than block
groups without high schools. Middle schools, however, appear to be unrelated to violent
crime at the block group level. These results are generally unsupportive of hypothesis 2
and 3, as there appears to be no significant relationship between the presence of
elementary schools or churches and violent crime at the block group level.
Table 4.1. Regression Results on Violent Crime
Variable
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Control
Model
0.354**
(0.045)
0.318**
(0.032)
0.009**
(0.002)
-0.021**
(0.004)
-
School
Model
0.366**
(0.044)
0.307**
(0.032)
0.009**
(0.002)
-0.020**
(0.004)
-0.018
(0.077)
0.200
(0.125)
0.345*
(0.166)
Church
Model
0.353**
(0.045)
0.318**
(0.032)
0.009**
(0.002)
-0.021**
(0.004)
Alcohol
Model
0.353**
(0.038)
0.242**
(0.028)
0.009**
(0.002)
-0.019**
(0.004)
-
-
-
-
-
-
Middle School
-
High School
-
Church
-
-
0.037
(0.060)
Bar
-
-
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
0.006**
(0.001)
-3.815**
(0.106)
0.0905
-1787.72
0.006**
(0.001)
-3.832**
(0.105)
0.0926
-1783.73
0.006**
(0.001)
-3.827**
(0.107)
0.0906
-1787.53
Lag
Constant
Pseudo R2
LL
0.418**
(0.065)
0.381***
(0.060)
0.092
(0.108)
0.255**
(0.059)
0.005**
(0.001)
-4.139**
(0.096)
0.1232
-1723.51
Complete
Model
0.363**
(0.038)
0.239**
(0.028)
0.009**
(0.002)
-0.018**
(0.004)
0.083
(0.068)
0.148
(0.108)
0.256
(0.143)
-0.073
(0.054)
0.425**
(0.065)
0.385**
(0.060)
0.097
(0.107)
0.262**
(0.059)
0.005**
(0.001)
-4.156**
(0.096)
0.1255
-1719.04
Standard errors in parentheses. * p < 0.05, ** p < 0.01
22
Log-likelihood ratio tests and the pseudo R-squared values confirm that alcohol-serving
and -selling establishments are more important predictors of violent crime incidents than
schools and churches. The pseudo R-squared value for the control model is 0.0905. This
value increases to 0.0926 for the school model and 0.0906 for the church model.
Conversely, the pseudo R-squared value for the alcohol model is 0.1232. Similarly, the
log-likelihood test for the alcohol model ( χ 2 = 128.42, χ 02 = 9.49 ) indicates that the
inclusion of places that serve and sell alcohol improves the fit of the control model at the
0.001 level of significance, while these tests suggest that schools ( χ 2 = 7.98, χ 02 = 7.82 )
only improve the fit of the control model at the 0.05 level of significance, and churches
( χ 2 = 0.38, χ 02 = 3.84 ) do not appear to significantly improve the fit of the control model
at any standard level of statistical significance. Similarly, the addition of schools and
churches to the alcohol model does not statistically significantly improve the alcohol
model’s fit ( χ 2 = 8.94, χ 02 = 9.49 ).
Place and Property Crime
Table 4.2 displays the regression results for property crime. In terms of the control
variables, many of these results are substantively similar to the results for violent crime.
Block groups with higher levels of disadvantage and instability have higher property
crime rates, while block groups with larger youth populations tend to have lower property
crime rates. The spatial lag variable is statistically significant for this model as well,
indicating that nearby block groups tend to have similar property crime rates, controlling
for other factors. A key difference between the property and violent crime models is that
the percentage of a block group’s population that is Hispanic is more strongly related to
violent crime than to property crime. While there is evidence of a positive statistically
significant relationship between percentage Hispanic and property crime, this relationship
is not significant in all of the property crime models.
In terms of place, there is a statistically significant relationship between schools and
property crime and alcohol-serving and -selling establishments and property crime.
Churches, however, are not significantly related to property crime. More specifically,
block groups with high schools are expected to have 96.4% ( e 0.675 = 1.964 ) more
property crime incidents than block groups without high schools, controlling for other
factors. There is also evidence that block groups with elementary schools are expected to
have significantly fewer property crimes than block groups without elementary schools,
though elementary schools are not a significant predictor of property crime in the
complete model. These results are generally supportive of our hypotheses regarding
schools and crime, as they indicate that block groups with high schools have, controlling
for other factors, more property crime, while block groups with elementary schools may,
controlling for other factors, report less property crime.
23
Bars, restaurants that serve beer and wine, and retail establishments that sell alcohol are
all significantly associated with elevated levels of property crime. Controlling for other
factors, block groups with bars are expected to have 62.9% ( e 0.488 = 1.629 ) more property
crime incidents than block groups without bars, block groups with retail establishments
that sell alcohol are expected to have 51.4% ( e 0.415 = 1.514 ) more property crime
incidents than block groups without alcohol-selling establishments, and block groups
with restaurants beer and wine are expected to have 28.5% ( e 0.251 = 1.285 ) more property
crime incidents than block groups without restaurants that serve beer and wine.
Table 4.2. Regression Results on Property Crime
Variable
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Control
Model
0.143**
(0.048)
0.260**
(0.035)
0.005*
(0.002)
-0.021**
(0.004)
-
School
Model
0.153**
(0.046)
0.221**
(0.034)
0.005*
(0.002)
-0.021**
(0.004)
-0.204*
(0.083)
0.032
(0.138)
0.826**
(0.183)
Church
Model
0.144**
(0.048)
0.246**
(0.035)
0.005*
(0.002)
-0.021**
(0.004)
Alcohol
Model
0.174**
(0.042)
0.156**
(0.031)
0.003
(0.002)
-0.017**
(0.004)
-
-
-
-
-
-
Middle School
-
High School
-
Church
-
-
0.127
(0.067)
Bar
-
-
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
0.001**
(<0.001)
-1.833**
(0.113)
0.0329
-2633.26
0.001**
(<0.001)
-.845**
(0.111)
0.0389
-2617.12
0.001**
(<0.001)
-1.870**
(0.114)
0.0336
-2631.47
Lag
Constant
Pseudo R2
LL
0.544**
(0.075)
0.417**
(0.068)
0.082
(0.123)
0.257**
(0.068)
0.001**
(<0.001)
-2.109**
(0.101)
0.0577
-2565.96
Complete
Model
0.177**
(0.041)
0.137**
(0.031)
0.003
(0.002)
-0.017**
(0.004)
-0.104
(0.074)
0.053
(0.121)
0.675**
(0.161)
0.009
(0.059)
0.488**
(0.074)
0.415**
(0.067)
0.072
(0.121)
0.251**
(0.067)
0.001**
(<0.001)
-2.122**
(0.100)
0.0619
-2554.29
Standard errors in parentheses. * p < 0.05, ** p < 0.01
24
These results are supportive of our hypotheses regarding alcohol serving and selling
establishments and crime in that block groups with alcohol-selling and -serving
establishments are expected to report significantly higher counts of property crime,
though this effect is larger for bars than for other types of alcohol-related establishments.
Interestingly, however, the effect of high school presence on property crime is larger than
the effect of bars on property crime.
Log-likelihood ratio tests and the pseudo R-squared values indicate that alcohol serving
and selling establishments are more important predictors of property crime incidents than
schools, which in turn are more important than churches. The difference in both pseudo
R-squared values and in log-likelihoods is larger for the alcohol model than it is for the
schools model. This result seems contrary to the result that the regression coefficient for
high schools is larger than the regression coefficient for any other type of place. This
apparent difference may reflect the fact that several different types of alcohol severing
and selling establishments are significantly related to crime in the complete model, while
high schools are the only type of school that is consistently related to property crime.
Place and Narcotics Crime
Table 4.3 displays the regression results for narcotics incidents. In terms of control
variables, these results are substantively very similar to the results for violent and
property crime. The spatial lag variable is again statistically significant across all
models, indicating that nearby block groups tend to be similar in terms of narcotics
incidence rates. Block groups with higher levels of disadvantage, instability, and greater
proportions of Hispanic residents report significantly more narcotics incidents than other
block groups. Conversely, block groups with larger proportions of youths report
significantly fewer narcotics incidents than other block groups.
In terms of place, there is again no statistically significant relationship between the
presence of a church and narcotics incidents at the block group level. Both schools and
alcohol-serving and -selling establishments are significantly related to narcotics incidents
at the block group level. Regarding schools, block groups with high schools are expected
to report 420.7% ( e1.650 = 5.207 ) more narcotics incidents than block groups without high
schools, while block groups with middle schools are expected to report 122.3%
( e 0.799 = 2.223 ) more narcotics incidents than block groups without middle schools.
Elementary schools are not significantly related to narcotics incidents. Block groups with
high schools and middle schools have elevated rates of narcotics incidents, though the
presence of a high school is a much larger risk factor than the presence of a middle
school. These results are generally supportive of our hypotheses regarding the
relationship between schools and neighborhood crime, though we do not observe any
protective effect of elementary schools on narcotics at the block group level.
25
Table 4.3. Regression Results on Narcotics Incidents
Variable
High School
-
Church
-
School
Model
0.531**
(0.063)
0.304**
(0.045)
0.012**
(0.003)
-0.030**
(0.005)
0.012
(0.105)
0.680**
(0.170)
1.564**
(0.225)
-
Bar
-
-
0.097
(0.090)
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
Lag
0.003**
(0.001)
-3.563**
(0.146)
0.004**
(0.001)
-3.799**
(0.135)
0.0732
-1882.91
0.0954
-1837.74
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Middle School
Constant
Pseudo R2
LL
Control
Model
0.495**
(0.070)
0.356**
(0.049)
0.011**
(0.003)
-0.030**
(0.005)
-
Church
Model
0.497**
(0.070)
0.349**
(0.050)
0.011**
(0.003)
-0.030**
(0.005)
-
Alcohol
Model
0.512**
(0.065)
0.257**
(0.047)
0.010**
(0.003)
-0.028**
(0.005)
-
0.003**
(0.001)
-3.589**
(0.148)
0.355**
(0.108)
0.349**
(0.099)
0.492**
(0.180)
0.254**
(0.095)
0.003**
(0.001)
-3.846**
(0.143)
Complete
Model
0.541**
(0.057)
0.209**
(0.041)
0.010**
(0.003)
-0.026**
(0.005)
0.092
(0.097)
0.799**
(0.155)
1.650**
(0.203)
0.032
(0.077)
0.325**
(0.096)
0.407**
(0.087)
0.536**
(0.158)
0.320**
(0.086)
0.004**
(0.001)
-4.183**
(0.129)
-
-
-
-
0.0735
-1882.34
0.0886
-1851.52
0.1182
-1791.56
Standard errors in parentheses. * p < 0.05, ** p < 0.01
All types of alcohol establishments are positively and significantly related to narcotics
incidents at the block group level. Controlling for other factors, block groups with bars
are expected to report 38.4% ( e 0.325 = 1.384 ) more narcotics incidents than other block
groups, block groups with retail establishments that sell alcohol are expected to report
50.2% ( e 0.407 = 1.502 ) more narcotics than other block groups, block groups with liquor
clubs are expected to report 70.9% ( e 0.536 = 1.709 ) more narcotics incidents than other
26
block groups, and block groups with restaurants that serve beer and wine are expected to
report 37.7% ( e 0.320 = 1.377 ) more narcotics incidents than other block groups. These
results are highly supportive of our general hypothesis regarding alcohol establishments
and crime, as block groups containing alcohol establishments are expected to report
significantly more narcotics incidents than block groups without alcohol establishments.
Our results, however, are not supportive of hypothesis 5 regarding the importance of bars.
In terms of narcotics incidents, both retail establishments that sell alcohol and liquor
clubs are a larger risk factor for narcotics incidents at the block group level than bars.
Interestingly, log-likelihood ratio tests and the pseudo R-squared values indicate that
schools are a more important predictor of narcotics incidents than alcohol establishments.
The difference in the pseudo R-squared values and in log-likelihoods is larger between
the alcohol model and control model than it is between the school and control model.
Moreover, the regression coefficient for high schools is the largest of all place variables
and the regression coefficient for middle schools is the second largest. While schools
may have a larger effect on narcotics incidents at the block group level, alcohol
establishments are still important. Each of the alcohol variables is a significant predictor
of narcotics incidents at the block group level and the log-likelihood test comparing the
school and complete models suggests that the inclusion of the alcohol establishment
variables significantly improves the school model’s fit ( χ 2 = 92.36, χ 02 = 11.07 ).
27
CHAPTER V: DISCUSSION AND CONCLUSION
The results of our regression analyses indicate that alcohol establishments and certain
types of schools are risk factors for crime at the block group level. Specifically, our
results indicate that block groups that contain alcohol establishments and, to a lesser
degree, block groups that contain high schools report more crime. It is important to note
that these results hold independent of the structural conditions of block groups. By
utilizing the block group level of analysis, we were able to include measures to
statistically control for structural disadvantage, residential instability, and demographic
factors. That there is still a strong, statistically significant relationship between different
types of places and crime, even after controlling for these factors, indicates that the
relationship between place and crime is not spurious by way of block group
characteristics.
In terms of schools, our regression results indicate that block groups with high schools
are likely to report more property and narcotics crime, controlling for other factors. In
fact, the regression coefficients for high schools are larger than the regression coefficients
for any other types of place, suggesting that high school presence is the strongest place
predictor of property and narcotics crime. Middle schools, though generally unrelated to
violent and property crime in our models, are the second strongest predictor of narcotics
incidents. These results support our first hypothesis, though it should be reiterated that
high schools are a much more salient predictor of neighborhood crime than middle
schools.
These results make sense in the context of the routine activities perspective. Individuals
in the high school age group are, as previously mentioned, at an increased risk to be both
the victims and offenders of property crime (Hirschi and Gottfredson, 1983; Farrington,
1986; Rand and Catalano, 2007). Given that people from these age groups converge in
large numbers in and around schools, it is unsurprising that areas with high schools have
elevated property crime rates. Nationwide, the Centers for Disease Control and
Prevention estimates that approximately 20% of teenagers admit to at least occasionally
using some illegal drug (Eaton et al., 2010), so it seems sensible that drug dealers would
attempt to converge with their potential clients in and around schools.
From a social disorganization perspective, our results may indicate that high schools, and
to a lesser degree, middle schools impede neighborhood collective efficacy processes.
Unfortunately, we do not have access to the type of data necessary to explore this
possibility. Future research should investigate the degree to which block groups with
high schools differ from other block groups in terms of key measures of social
organization (e.g., participation in local community groups and social network measures).
In terms of alcohol establishments, bars, restaurants that serve beer and wine, and retail
establishments that sell alcohol are significant predictors of violent, property, and
28
narcotics crime at the block group level. Liquor clubs, conversely, are only significantly
related to narcotics offenses. Narcotics incidents are not associated with specific victims.
This might suggest that while liquor clubs do promote the convergence of people in time
and space, these clubs are not promoting the convergence of people that are likely to
victimize each other.
The other types of alcohol establishments are associated with elevated levels of violent,
property, and drug crime. This finding clearly fits within the framework of the routine
activities perspective, as each of these types of places brings together large groups of
people. Bars are the largest place predictor of violent crime at the block group level and
the second largest place predictor of property crime at the block group level. These
findings generally support our fifth hypothesis and indicate that bars have a greater
criminogenic effect than other alcohol establishments. Unfortunately, we are unable to
specify exactly why bars are such a significant risk factor for crime at the block group
level. It may be the case, as Roncek and Maier (1991) have previously argued, that the
consumption of alcohol at bars increases both the average criminal motivation and target
attractiveness of bar patrons. Alternatively, the relationship between bars and
neighborhood crime may be reflective of some selection effect in that people who are
likely to engage in and be the victims of crime are simply more likely to frequent bars
than other people. Bars may also interfere with collective efficacy processes by impeding
the formation or reducing the strength of neighborhood-level informal social controls.
Local patrons may be less beholden to neighborhood-level social controls when under the
influence of alcohol. Additionally, local social controls likely hold little sway over
patrons from outside the neighborhood.
Interestingly, retail establishments that sell alcohol are also significantly related to crime
at the block group level. While the retail establishments include places that would be
traditionally defined as liquor stores, the majority of places that possess this type of
liquor license in Albuquerque are better viewed as big box and grocery stores. Clearly,
the above arguments regarding bars and neighborhood crime do not cleanly apply to big
box stores or grocery stores. It is possible that the relationship between retail
establishments that sell alcohol and neighborhood crime is simply reflective of the sheer
volume of people that converge in and around these locations, many of whom may come
from outside the neighborhood. Future research should compare the criminogenic
properties of retail establishments that serve alcohol to those that do not.
Our results indicate that there is no statistically significant relationship between the
presence of religious places and crime at the block group level. This result contradicts
our hypothesis regarding religious places and crime at the block group level. Based on
the routine activities and social disorganization traditions, we had expected religious
places to have a negative relationship with block group crime, as we believed that
churches might promote routine activity patterns that are not conducive to crime, while
simultaneously promoting the formation of pro-social friendship networks. While we
were unable to find any evidence of a protective role for churches, we believe that the
potential relationship between churches and neighborhood crime deserves additional
29
study. The measure of religious places that we employed in our analysis (the presence or
absence of a place of worship in a block group) may be too simplistic to evaluate the role
of churches in terms of neighborhood crime.
The presence of a religious place does not indicate the usage of that place within a
neighborhood. Certain religious places may be busier than other churches and this level
of usage may be reflective of how effective a religious place can be in terms of the
promotion of social ties. Some religious places may, for example, be more heavily
involved in their neighborhoods than others. Further, some religious places, especially
those that serve homeless and felon populations, may in fact increase crime at the
neighborhood level by promoting the convergence of these populations in time and space.
Perhaps more importantly, our measure of religious places cannot distinguish between
different types of religious places. Our measure simply indicates whether or not there is
at least one religious place within a block group, but it does not indicate the religious
affiliation of these religious places within a block group. While it is true that most of the
churches in Albuquerque are Christian in nature, our measure is also unable to distinguish
between different types of Christian churches. DiIolio (2009: 129), for example, notes
that most theory and research on religion and crime approaches the issue from a
“churched/not churched” perspective. He notes, however, that there are several types of
religious organizations and suggests that these organizations may be differentially related
to crime. It is important for future research on religious places and neighborhood crime
to distinguish between different types of religious places and, if possible, to incorporate
data on the use of these places (both in terms of frequency and types of services offered).
In general, our results suggest that specific types of places are risk factors for
neighborhood crime.
Elementary schools are the only type of place included in our study that has a statistically
significant negative relationship with property crime at the block group level. This result,
however, is not as strong as the positive relationships described above (elementary
schools are significant in the school model, but not in the complete model). Given that
we controlled for a wide variety of other factors, this result may suggest that the presence
of elementary schools may reduce property crime. Both the social disorganization and
routine activities perspectives provide arguments that might explain the protective role of
elementary schools. Unlike middle schools and high schools, elementary schools do not
funnel large groups of people in neighborhoods that, on average, are more likely to be
criminal offenders and victims. In other words, elementary school students are not likely
to engage in property crime, nor are they likely to be the victims of property offenses.
Further, the smaller student-to-teacher ratio and higher degree of parental involvement in
elementary schools may foster both a heightened level of guardianship and the formation
of positive social ties, both of which may serve to hinder property crime.
Interestingly, the presence of elementary schools is not associated with lower levels of
violent crime or drug crime. Controlling for other factors, there is no statistically
significant relationship between elementary school presence and these types of crimes,
30
suggesting that block groups with elementary schools are no different than other block
groups in terms of violent crime or drug crime. We are unable to state why elementary
schools seem to protect block groups against property crime and not violent or drug
crime. The routine activities perspective would suggest that elementary schools promote
routine activity patterns that are discouraging to the commission of property crime and
that are neither conducive nor discouraging to violent crime and drug crime. Similarly,
the social disorganization perspective might argue that elementary schools produce
informal social control that is effective at deterring property crime but not violent or drug
crime.
It may be the case, for example, that elementary schools promote a specific style of
guardianship and a specific style of social control, both focused on the monitoring of
elementary school students. That is, perhaps the protective role of elementary schools is
directly related to the presence and visual supervision of teachers and parents. The
presence of these individuals may deter would-be burglars, as it may be difficult to break
into homes in these neighborhoods. On the other hand, drug crimes and the serious sort
of violent crimes that we considered in our analysis may be more likely to occur behind
closed doors and are thus less likely to be deterred by the presence of teachers and
parents. Moreover, if elementary schools primarily promote social organization for the
monitoring and protection of elementary school-aged children, it may be the case that
elementary schools do not foster the types of social bonds that are likely to deter other
types of crime.
Future research on place and neighborhood crime should attempt to address this issue and
more completely describe the processes through which specific places influence
neighborhood crime. For example, it would be useful to better understand how
neighborhoods with and without elementary schools vary in terms of routine activity
patterns and overall social organization. If, as we speculated above, elementary schools
promote a type of social control based on the supervision of public spaces, we might
expect neighborhoods with elementary schools to have more active neighborhood
watches and neighborhood association groups and yet not necessarily denser pro-social
friendship networks.
While we have advocated for a more thorough examination of how elementary schools
shape neighborhood activity patterns and neighborhood organization, we believe that this
approach can improve the criminological knowledge on place and crime more broadly.
To date, most research on place and crime, including the current study, has simply
demonstrated that there is an empirical correlation between the presence of certain types
of places and neighborhood crime rates. This research, while useful, is only a first step
toward understanding the relationship between place and crime. Beyond describing
which types of places are, on average, positively or negative related to crime, future
research, utilizing neighborhood surveys and qualitative data, should focus on describing
how these places are related to crime. This knowledge may both highlight the
conditional scopes of theoretical traditions and be useful in the development of placespecific crime prevention strategies.
31
These results highlight several policy implications on neighborhood crime. First, given
the robust and highly significant relationship between the presence of bars and all types
of crime, we believe that public policy makers should focus special attention on
neighborhoods containing bars. The routine activities perspective suggests that these
areas are likely to see elevated crime rates due to the fact that there is not enough capable
guardianship in and around bars to counter the convergence of the large volume of
would-be victims and offenders. The obvious policy implications are to either increase
levels of guardianship (increased security in and around bars and increased police
presence in these neighborhoods) or to decrease the number of motivated offenders and
suitable victims (this may be possible, for example, by imposing capacity limitations in
bars based on crime instead of fire considerations). Other target hardening strategies, like
increased street lighting and the placement of security cameras may also be useful in
reducing crime in and around bars. Furthermore, neighborhood associations and related
watch groups may be able to provide important support for these efforts, by extending
their watch and influence to both public and commercial areas. Many of these strategies
may also be useful in addressing the moderately strong relationship between the presence
of restaurants that serve alcohol and crime.
Similarly, these results suggest that additional attention to neighborhoods with high
schools, and to a lesser extent, to neighborhoods with middle schools is warranted. Many
of the policy implications for neighborhoods with bars also apply to neighborhoods with
schools. These neighborhoods may benefit from an increased presence of teachers,
school security, and police officers in and around schools, especially during the
commutes to and from schools. Neighborhood associations and watch groups near high
schools may also be able to reduce the criminogenic effect of these schools by carefully
scheduling watch activities before and after school in the areas surrounding the schools.
These strategies would expand the scope of guardianship to include both the school
grounds and the surrounding areas. We believe that it is important for future research to
study specific situational crime strategies and neighborhood organizational patterns to
better develop a list of which strategies and approaches are most effective at reducing the
criminogenic effects of specific types of places.
We conclude by noting that our research has several limitations that must be mentioned.
While we believe that our results are explainable from the routine activities and social
disorganization perspectives, we do not have actual measures of these processes. Further,
we have focused largely on serious offenses as we do not have access to geocoded data
for less serious types of crimes. It may be the case that our results would not apply to
less serious violent and property offenses. Future research should investigate the
relationship between place and less serious forms of crime.
Despite these limitations, we feel that the current research is valuable in many ways.
This research has demonstrated that alcohol establishments and schools are associated
with certain types of crime at the block group level, even after controlling for a number
of factors that are known to be associated with crime. Most of the previous research on
32
place and crime has been either completely descriptive (calculating the proportion of
incidents that happen at specific places) or has been conducted at the block level of
analysis (which precludes controlling for a wide range of social structural variables).
33
REFERENCES
Alvarez-Rivera, Lorna L., Fox, Kathleen A. (2010). Institutional attachments and selfcontrol: Understanding deviance among Hispanic adolescents. Journal of Criminal
Justice 38,666-674.
Anselin, L. (1992). Spatial data analysis with GIS: An introduction to application in the
social sciences. Santa Barbara, CA: National Center for Geographic Information and
Analysis, University of California.
Block, R. and Block C.R. (1995). Space, place, and crime: Hot spots areas and hot places
of liquor-related crime. In J. Eck and D. Weisburd, (Eds.), Crime and place (pp. 145–
183). Monsey, NY: Criminal Justice Press.
Broidy, L., Willits, D., and Denman, K. (2009). Schools and neighborhood crime (Final
Report to the Justice Research Statistics Association). Albuquerque: New Mexico
Statistical Analysis Center, Institute for Social Research, University of New Mexico.
Bursik, R. (1988). Social disorganization and theories of crime and delinquency:
Problems and prospects. Criminology, 26, 519–552.
Cohen, L. E. and Felson, M. (1979). Social change and crime rate trends: A routine
activity approach. American Sociological Review, 44, 588–608.
DiIolio, J. (2009). More religion, less crime? Science, felonies, and the three faith factors.
Annual Review of Law and Social Science, 5, 115–133.
Dobson, A. J. (2002). An introduction to generalized linear models. Florida: Chapman &
Hall/CRC Press.
Dunteman, G. (1989). Principal components analysis. Newbury Park, California: Sage.
Eaton, D., Kann, L., Kinchen, S., Shanklin, S., Ross, J., Hawkins, J., Harris, W., Lowry,
R, McManus, T., Chyen, D., Lim, C., Whittle, L., Brener, N., and Wechsler H. (2010).
Youth risk behavior surveillance – United States, 2009. Surveillance Summaries, 59, 1142.
Eccles, J. S. and Harold, R. D. (1996). Family involvement in children’s and adolscents’
schooling. In J. D. A Booth (Ed.), Family-school links: How do they affect educational
outcomes (pp 3–34). Hillsdale, NJ: Lawrence Erlbaum Associates.
Farrington, D. (1986). Age and crime. Crime and Justice, 7, 189–250.
34
Garofalo, J., Siegel, L., and Laub, J. (1987). School-related victimizations among
adolescents: An analysis of National Crime Survey (NCS) narratives. Journal of
Quantitative Criminology, 3, 321–338.
Hill, N. E. and Taylor, L. C. (2004). Parental school involvement and children’s
academic achievement: Pragmatics and issues. Current Directions in Psychological
Science, 13, 161–164.
Hirschi, T., and Gottfredson, M. (1983). Age and the explanation of crime. American
Journal of Sociology 89, 552-584.
Ito, T. A., Miller, N., and Pollock, V. E. (1996). Alcohol and aggression: A meta-analysis
on the moderating effects of inhibitory cues, triggering events, and self-focused attention.
Psychological Bulletin, 120, 60–82.
Johnson, Byron R., Jang, Sung Joon, De Li, Spencer, Larson, David. (2000). The
‘Invisible Institution’ and Black Youth Crime: The Church as an Agency of Local Social
Control. Journal of Youth and Adolescence 29, 479-498
Kasarda, J. and Janowitz, M. (1974). Community attachment in mass society. American
Sociological Review, 39, 28–39
Kautt, P. M. and Roncek, D. W. (2007). Schools as criminal “hot spots”: Primary,
secondary, and beyond. Criminal Justice Review, 32, 339–357.
Knudten, Richard D., Knudten, Mary S. (1971). Juvenile Delinquency, Crime, and
Religion. Review of Religious Research 12,130-152
Krivo, L., and Peterson, R. (1996). Extremely disadvantaged neighborhoods and urban
crime. Social Forces, 75, 619-650.
Lee, M. (2006). The religious institutional base and violent crime in rural areas. Journal
for the Scientific Study of Religion, 45, 309–324.
Lee, M. 2008. Civic community in the hinterland: Toward a theory of rural social
structure and violence. Criminology, 46, 447–478.
Lee, M. (2010). Civic community, population change, and violent crime in rural
communities. Journal of Research in Crime and Delinquency, 47, 118–147.
Mosher, C. J., Miethe, T. D., and Phillips, D. M. (2002). The mismeasure of crime.
Thousand Oaks, CA: Sage Publications.
35
Murray, R. and Swatt, M. (2010). Disaggregating the relationship between schools and
crime: A spatial analysis. Crime & Deliqnuency. Available online <
http://cad.sagepub.com/content/early/2010/01/05/0011128709348438>.
Osgood, D. W. (2000). Poisson-based regression analysis of aggregate crime rates.
Journal of Quantitative Criminology, 16, 21–43.
Peterson, R., Krivo, L., and Harris, M.. (2000). Disadvantage and neighborhood violent
crime: Do local institutions matter? Journal of Research in Crime and Delinquency, 37,
31–63.
Rand, M. and Catalano, S. (2007). Criminal victimization, 2006. Washington, DC: U.S.
Department of Justice.
Roman, C.G. (2004). Schools, neighborhoods, and violence: Crime within the daily
routines of youth. Lanham, MD: Lexington Books.
Roncek, D.W. (1981). Dangerous places: Crime and residential environment. Social
Forces, 60, 74–96.
Roncek, D.W. and Bell, R. (1981). Bars, blocks, and crime. Journal of Environmental
Systems, 11, 35–47.
Roncek, D.W. and Faggiani, D. (1985). High schools and crime: A replication.
Sociological Quarterly, 26, 491–505.
Roncek, D.W. and Lobosco, A. (1983). Effect of high schools on crime in their
neighborhoods. Social Science Quarterly, 64(3), 598–613.
Roncek, D.W. and Maier, P.A. (1991). Bars, blocks, and crimes revisited: Linking the
theory of routine activities to the empiricism of “hot spots.” Criminology, 29, 725–753.
Roncek, D. W. and Pravatiner, M. A. (1989). Additional evidence that taverns enhance
nearby crime. Sociology and Social Research, 73, 185–188.
Rose, Dina R., Clear, Todd R. (2006). Incarceration, Social Capital, and Crime:
Implications for Social Disorganization Theory. Criminology 36, 441-480
Sampson, R. J. and Groves, W. B. (1989). Community structure and crime: Testing
social-disorganization theory. American Journal of Sociology, 94, 774–802.
Sampson, R. J. and Wilson,W. J. (1995). Toward a theory of race, crime, and urban
inequality. In J. Hagon and R. D. Peterson (Eds.), Crime and inequality (pp. 37–54).
Stanford, CA: Stanford University Press.
36
Sampson, R. J., Morenoff, J. D., and Gannon-Rowley, T. (2002). Assessing
“neighborhood effects”: Social processes and new directions in research. Annual Review
of Sociology, 28, 443–478.
Sampson, R. J., Raudenbush, S. W., and Earls, F. (1997). Neighborhoods and violent
crime: A multilevel study of collective efficacy. Science, 277, 918–924.
Shaw, C. and McKay, H. (1942). Juvenile delinquency and urban areas. Chicago:
University of Chicago Press.
Sherman, L. W., Gartin, P. R., and Buerger, M. E. (1989). Hot spots of predatory crime:
Routine activities and the criminology of place. Criminology, 27(1): 27– 56.
Spelman, W. (1995). Criminal careers of public places. In J. Eck and D. Weisburd (Eds.),
Crime and place (pp. 115–144). Monsey, NY: Criminal Justice Press.
Taylor, Robert Joseph, Thornton, Michael C., Chatters, Linda M. (1988). Black
Americans’ Perceptions of the Sociohistorical Role of the Church. Journal of Black
Studies 18, 123-138
Taylor, Robert Joseph, Chatters, Linda M. (1988). Church Members as a Source of
Informal Social Support. Review of Religious Research 30, 193-203.
Wimberly, E. (1979). Pastoral Care in the Black Church. Nashville: Adingdon.
37
APPENDIX: ADDITIONAL REGRESSION RESULTS
Table A1. Poisson Regression Results Using Homicide as the
Dependent Variable
Variable
Church
-
School
Model
0.451**
(0.076)
0.467**
(0.067)
0.020**
(0.004)
-0.047**
(0.009)
0.085
(0.170)
0.005
(0.273)
-0.393
(0.350)
-
Bar
-
-
0.268
(0.139)
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
Constant
-7.691**
(0.227)
0.2051
-380.40
-7.724**
(0.232)
0.2069
-379.56
-7.790**
(0.232)
0.2092
-378.77
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Middle
School
High School
Pseudo R2
LL
Control
Model
0.466**
(0.074)
0.448**
(0.064)
0.019**
(0.004)
-0.047**
(0.009)
-
Church
Model
0.452**
(0.074)
0.441**
(0.065)
0.018**
0.004)
-0.047**
(0.009)
-
Alcohol
Model
0.426**
(0.076)
0.375**
(0.070)
0.022**
(0.004)
-0.046**
(0.010)
-
-
-
-
0.225
(0.150)
0.527**
(0.140)
0.400*
(0.188)
0.150
(0.144)
-8.171**
(0.255)
0.2376
-364.85
Complete
Model
0.395**
(0.079)
0.404**
(0.072)
0.023**
(0.004)
-0.045**
(0.010)
0.189
(0.176)
0.001
(0.281)
-0.566
(0.365)
0.109
(0.146)
0.268
(0.154)
0.559**
(0.142)
0.360
(0.193)
0.117
(0.148)
-8.302**
(0.269)
0.2425
-362.52
Standard errors in parentheses. * p < 0.05, ** p < 0.01
38
Table A2. Negative Binomial Regression Results Using Rape as the
Dependent Variable
Variable
Church
-
School
Model
0.277**
(0.054)
0.418**
(0.040)
0.011**
(0.003)
-0.020**
(0.005)
0.074
(0.102)
0.327*
(0.153)
0.228
(0.193)
-
Bar
-
-
-0.017
(0.081)
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
Constant
-6.032**
(0.139)
0.0996
-906.70
-6.061**
(0.139)
0.1030
-903.29
-6.027**
(0.0141)
0.0996
-906.68
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Middle
School
High School
Pseudo R2
LL
Control
Model
0.252**
(0.054)
0.422**
(0.039)
0.012**
(0.003)
-0.021**
(0.005)
-
Church
Model
0.253**
(0.054)
0.422**
(0.039)
0.012**
(0.003)
-0.021**
(0.005)
-
Alcohol
Model
0.222**
(0.053)
0.378**
(0.040)
0.012**
(0.003)
-0.017**
(0.006)
-
-
-
-
0.256**
(0.095)
0.092
(0.089)
0.256
(0.148)
0.091
(0.087)
-6.267**
(0.148)
0.1089
-897.32
Complete
Model
0.274**
(0.053)
0.380**
(0.040)
0.012**
(0.003)
-0.016**
(0.006)
0.121
(0.101)
0.320*
(0.167)
0.167
(0.189)
-0.100
(0.082)
0.277**
(0.095)
0.101
(0.089)
0.280
(0.146)
0.087
(0.088)
-6.289**
(0.148)
0.1133
-892.92
Standard errors in parentheses. * p < 0.05, ** p < 0.01
39
Table A3. Negative Binomial Regression Results Using Aggravated
Assault as the Dependent Variable
Variable
Church
-
School
Model
0.373**
(0.040)
0.336**
(0.029)
0.011**
(0.002)
-0.019**
(0.004)
-0.020
(0.069)
0.227*
(0.111)
0.357*
(0.148)
-
Bar
-
-
0.059
(0.054)
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
Lag
0.004**
(0.001)
-3.827**
(0.094)
0.1102
-1780.41
0.004**
(0.001)
-3.841**
(0.094)
0.1131
-1774.45
0.004**
(0.001)
-3.843**
(0.095)
0.1105
-1779.80
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Middle
School
High School
Constant
Pseudo R2
LL
Control
Model
0.357**
(0.040)
0.358**
(0.029)
0.011**
(0.002)
-0.019**
(0.004)
-
Church
Model
0.356**
(0.040)
0.346**
(0.029)
0.011**
(0.002)
-0.019**
(0.004)
-
Alcohol
Model
0.352**
(0.035)
0.286**
(0.026)
0.011**
(0.002)
-0.016**
(0.003)
-
-
-
-
0.347**
(0.060)
0.246**
(0.056)
0.160
(0.100)
0.187**
(0.055)
0.004**
(0.001)
-4.082**
(0.089)
0.1333
-1734.16
Complete
Model
0.366**
(0.035)
0.281**
(0.026)
0.011**
(0.002)
-0.016**
(0.003)
0.050
(0.063)
0.196*
(0.100)
0.285*
(0.133)
-0.032
(0.050)
0.345**
(0.061)
0.247**
(0.056)
0.162
(0.100)
0.188**
(0.056)
0.003**
(0.001)
-4.099**
(0.089)
0.1359
-1728.91
Standard errors in parentheses. * p < 0.05, ** p < 0.01
40
Table A4. Negative Binomial Regression Results Using Robbery as
the Dependent Variable
Variable
Church
-
School
Model
0.340**
(0.079)
0.263**
(0.054)
0.004
(0.003)
-0.024**
(0.007)
-0.019
(0.132)
0.232
(0.212)
0.301)
(0.281)
-
Bar
-
-
0.075
(0.100)
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
Lag
0.035**
(0.005)
-4.750**
(0.192)
0.0572
-1533.29
0.036**
(0.005)
-4.762**
(0.192)
0.0581
-1531.92
0.035**
(0.005)
-4.779**
(0.195)
0.0574
-1533.01
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Middle
School
High School
Constant
Pseudo R2
LL
Control
Model
0.327**
(0.079)
0.276**
(0.053)
0.004
(0.003)
-0.024**
(0.007)
-
Church
Model
0.325**
(0.079)
0.274**
(0.053)
0.004
(0.003)
-0.024**
(0.007)
-
Alcohol
Model
0.388**
(0.065)
0.141**
(0.045)
0.004
(0.003)
-0.021**
(0.006)
-
-
-
-
0.709**
(0.104)
0.810**
(0.095)
0.134
(0.172)
0.573**
(0.093)
0.026**
(0.004)
-5.323**
(0.154)
0.1117
-1444.72
Complete
Model
0.395**
(0.065)
0.139**
(0.045)
0.004
(0.003)
-0.021**
(0.006)
0.190
(0.108)
0.167
(0.175)
0.138
(0.228)
-0.124
(0.086)
0.734**
(0.105)
0.828**
(0.094)
0.152
(0.171)
0.585**
(0.093)
0.027**
(0.004)
-5.351**
(0.165)
0.1138
-1441.34
Standard errors in parentheses. * p < 0.05, ** p < 0.01
41
Table A5. Negative Binomial Regression Results Using Burglary as
the Dependent Variable
Variable
Church
-
School
Model
0.190**
(0.039)
0.132**
(0.028)
0.002
(0.002)
-0.015**
(0.003)
-0.140*
(0.066)
0.023
(0.109)
0.160
(0.143)
-
Bar
-
-
0.151**
(0.051)
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
Lag
0.006**
(0.001)
-3.087**
(0.101)
0.0528
-2093.62
0.007*
(0.001)
-3.077**
(0.101)
0.0541
-2090.71
0.006**
(0.001)
-3.124**
(0.100)
0.0548
-2089.15
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Middle
School
High School
Constant
Pseudo R2
LL
Control
Model
0.188**
(0.039)
0.144**
(0.028)
0.002
(0.002)
-0.016**
(0.003)
-
Church
Model
0.183**
(0.039)
0.139**
(0.027)
0.002
(0.002)
-0.016**
(0.003)
-
Alcohol
Model
0.185**
(0.037)
0.105**
(0.026)
0.002
(0.002)
-0.012**
(0.003)
-
-
-
-
0.294**
(0.060)
0.125*
(0.056)
0.120
(0.100)
0.1777
(0.055)
0.006**
(0.001)
-3.270**
(0.096)
0.0661
-2064.28
Complete
Model
0.183**
(0.037)
0.096**
(0.026)
0.002
(0.002)
-0.013**
(0.003)
-0.097
(0.062
0.014
(0.101)
0.067
(0.134)
0.097*
(0.049)
0.270**
(0.060)
0.121*
(0.056)
0.106
(0.099)
0.167**
(0.055)
0.006**
(0.001)
-3.270**
(0.096)
0.0675
-2061.17
Standard errors in parentheses. * p < 0.05, ** p < 0.01
42
Table A6. Negative Binomial Regression Results Using Larceny as
the Dependent Variable
Variable
Church
-
School
Model
0.151**
(0.052)
0.251**
(0.039)
0.004**
(0.003)
-0.024**
(0.004)
-0.237*
(0.095)
0.066
(0.157)
1.016**
(0.209)
-
Bar
-
-
0.137
(0.078)
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
Lag
0.001**
(0.0001)
-1.593**
(0.126)
0.0304
-2717.03
0.001**
(0.0001)
-1.604**
(0.123)
0.0373
-2697.59
0.001**
(0.0001)
-1.637**
(0.127)
0.0309
-2715.47
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Middle
School
High School
Constant
Pseudo R2
LL
Control
Model
0.137*
(0.055)
0.299**
(0.040)
0.004
(0.003)
-0.024**
(0.004)
-
Church
Model
0.138*
(0.054)
0.283**
(0.041)
0.003
(0.002)
-0.024**
(0.004)
-
Alcohol
Model
0.183**
(0.047)
0.162**
(0.035)
0.001
(0.002)
-0.020**
(0.004)
-
-
-
-
0.649**
(0.084)
0.580**
(0.075)
0.115
(0.138)
0.306**
(0.075)
0.001**
(0.0002)
-1.921**
(0.111)
0.0598
-2634.51
Complete
Model
0.185**
(0.045)
0.138**
(0.034)
0.001
(0.002)
-0.020**
(0.004)
-0.110
(0.081)
0.109
(0.134)
0.879**
(0.178)
-0.007
(0.0.065)
0.576**
(0.082)
0.590**
(0.074)
0.125
(0.134)
0.292**
(0.074)
0.0001**
(0.0001)
-1.943**
(0.108)
0.0658
-2617.80
Standard errors in parentheses. * p < 0.05, ** p < 0.01
43
Table A7. Negative Binomial Regression Results Using Motor Vehicle
Theft as the Dependent Variable
Variable
Church
-
School
Model
0.268**
(0.042)
0.272**
(0.031)
0.002
(0.002)
-0.014**
(0.004)
-0.123
(0.074)
0.132
(0.121)
0.229
(0.161)
-
Bar
-
-
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
Lag
0.008**
(0.001)
-3.504**
(0.103)
0.0732
-1953.99
0.008**
(0.001)
-3.493**
(0.104)
0.0747
-1950.88
0.008**
(0.001)
-3.518**
(0.105)
0.0734
-1953.70
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Middle
School
High School
Constant
Pseudo R2
LL
Control
Model
0.261**
(0.042)
0.287**
(0.031)
0.002
(0.002)
-0.015**
(0.004)
-
Church
Model
0.259**
(0.042)
0.285**
(0.031)
0.002
(0.002)
-0.015**
(0.004)
-
Alcohol
Model
0.257**
(0.038)
0.228**
(0.028)
0.001
(0.002)
-0.011**
(0.003)
-
-
-
-
-
0.044
(0.058)
-
0.453**
(0.065)
0.200**
(0.061)
0.083
(0.109)
0.183**
(0.060)
0.007**
(0.001)
-3.767**
(0.097)
0.0951
-1907.84
Complete
Model
0.264**
(0.038)
0.221**
(0.029)
0.001
(0.002)
-0.010**
(0.003)
-0.049
(0.067)
0.123
(0.109)
0.135
(0.144)
-0.045
(0.053)
(0.048)
0.453**
(0.065)
0.195**
(0.061)
0.084
(0.109)
0.187**
(0.060)
0.008**
(0.001)
-3.761**
(0.098)
0.0960
-1906.01
Standard errors in parentheses. * p < 0.05, ** p < 0.01
44
Table A8. Negative Binomial Regression Results Using Narcotics
Incidents as the Dependent Variable
Variable
Church
-
School
Model
0.531**
(0.063)
0.304**
(0.045)
0.012**
(0.003)
-0.030**
(0.005)
0.012
(0.105)
0.680**
(0.170)
1.564**
(0.225)
-
Bar
-
-
0.097
(0.090)
-
Retail
-
-
-
Liquor Club
-
-
-
Restaurant
-
-
-
Lag
0.003**
(0.001)
-3.563**
(0.146)
0.0732
-1882.91
0.004**
(0.001)
-3.799**
(0.135)
0.0954
-1837.74
0.003**
(0.001)
-3.589**
(0.148)
0.0735
-1882.34
Disadvantage
Instability
% Hispanic
% Under Age
18
Elementary
School
Middle
School
High School
Constant
Pseudo R2
LL
Control
Model
0.495**
(0.070)
0.356**
(0.049)
0.011**
(0.003)
-0.030**
(0.005)
-
Church
Model
0.497**
(0.070)
0.349**
(0.050)
0.011**
(0.003)
-0.030**
(0.005)
-
Alcohol
Model
0.512**
(0.065)
0.257**
(0.047)
0.010**
(0.003)
-0.028**
(0.005)
-
-
-
-
0.355**
(0.108)
0.349**
(0.099)
0.492**
(0.180)
0.254**
(0.095)
0.003**
(0.001)
-3.846**
(0.143)
0.0886
-1851.52
Complete
Model
0.541**
(0.057)
0.209**
(0.041)
0.010**
(0.003)
-0.026**
(0.005)
0.092
(0.097)
0.799**
(0.155)
1.650**
(0.203)
0.032
(0.077)
0.325**
(0.096)
0.407**
(0.087)
0.536**
(0.158)
0.320**
(0.086)
0.004**
(0.001)
-4.183**
(0.129)
0.1182
-1791.56
Standard errors in parentheses. * p < 0.05, ** p < 0.01
45